Hi,

I've spend few hours playing with the Bing Image Creator on the web and I must say the results are quite impressive with the right keywords used for the image creation. It's not perfect, the thing is still learning, but I genuinely like it and got addicted to generate any idea that comes to mind randomly sometimes.

I do some photography for fun but I have zero ability when it come to digital arts. Just want to hear some thoughts about the AI Image Creator for the members here. Is it good enough for you ? Is it good enough for you to use the picture for your music (or sample pack, whatever) releases covers ?

If possible please share some of your result here for fun. How would you like your next album/ep/release to look like and does the AI manages to get it right, or maybe close ? Anything.

Here's are some of the pictures generated by the AI that I think album cover worthy for various genres or maybe sample pack cover ? got tons of them but only manage to upload 5 here :















the avatar I'm using right now also AI generated graphic. I use the keyword "Moog Synthesizer Vector Art with Picasso style background". Not exactly a moog or a proper synth but yeah it's pretty and i like it.

 

Last edited: Apr 8, 2023

There's a bunch of free tools here - you can then use an upscaler and make the end result bigger: https://replicate.com/explore

 

interesting, need to learn about this stuff. thanks.

 

that's my tour photos done and dusted then

 

Heck yeah, seems like you could get something very usable out of that. Visual art seems to be AI's forte at the moment.

 

@Stevie Dude What prompts did you use for the first and fourth picture? They look really interesting
Also @arbuckle I´m interested yours prompts too
I tried these A.I´s year ago and I didn´t get anything close to these

 

Interesting. Is it good enough? I think so, but i also think AI generated images are still kind of recognizable as AI generated most of the time, something about the colour and overall surreal vibe that's more or less evident.

That's not a bad thing, but it may not be suitable for all scenarios.

I personally wouldn't use it because i don't feel the results go well with anything i do, but it's definetively usable. I've even seen some sample packs on sister site that use obviously AI generated art for the cover, so it's a thing.

 

If i recall correctly :

1. angry robot human internal bodyparts in the style of steampunk
4. apocalypse, city destroyed, steel on the ground, kite on the sky, sad, dark, black and white

there were like a lot of build-ups before I added those words. it was based on the previous result, so I tweak the query and add the things i want one by one.

similar to number 4 but i randomly added "ice cream truck" and "rainy" to it and it returns with this :



similar to number 1


this when you add "sunset"



changed "steel on the ground" to "dead cyborg on the ground"

 

I can't view your pictures. why others can see it ? hmm

you can crop or remove it with some editing. pretty sure there are a lot of other watermark in the file too, it was AI generated afterall but it doesn't really matter I guess.

 

Last edited: Apr 4, 2023

https://www.bing.com/images/create/retrofuturistic2c-fractal-walls2c-skull2c-wizard2c-cav/642c8ef0b0224b0bb74e46a33213e625?id=z0XLtb/EMYE3SKUPUbRsDw==&view=detailv2&idpp=genimg&FORM=GCRIDP
quick test

Couldn´t get the image on the message, so that´s why the link

 

It's better than the graph artists from my city...

 

pretty good at logos too although it blocks BAD WORDS like audiosex (but at least excited eggplants are cool)

 

So what? Is something wrong with it?

 

If someone is wondering: a AI generated imaged can't be copyrighted, not even by you or the AI machine owner. But if add or modify something from the content that is generated by the AI you can register as yours.

 

Pretty cool. Has anyone got the AI to correctly write something for you? It keeps spitting out something that looks like jumbled up classical Latin script for me.

 

It's actually better than other..
Has a better understanding of Language,
can create different objects in the same image without blending them into one as much..

 

Last edited: Apr 5, 2023

Couple more

 

Just the way it is at the moment, large language models like chatgpt (which bing uses for its chatbot) use mainly transformers that are trained to vectorise text and map a semantic latent space, whereas image generators like this or midjourney are more "traditional" convolutional neural networks that recognise the shape of words and letters but not the meaning, both are progressing really fast and there is some commingling of model architectures (and specialised models for reverse kcr and image <> text) but it hasn't happened yet in a super convincing way. Maybe by next Thursday?